Re: [gamin] Re: Runtime selectable backends for Gamin [u]

[Daniel Veillard <veillard redhat com>]

On Wed, Aug 25, 2004 at 09:22:57PM +0200, Martin Schlemmer [c] wrote:
> > > the code is using dnotify and polling at the same time now, so it's
> > > more complex than what your patch suggests.
> > > 
> > 
> > I should admit that it is not the most elegant solution, and will
> > not scale if more backends are used (will there be much more added
> > or in pipe-line to be added?).  I however wanted an easy way to
> > integrate both dnotify and inotify support without recompiling,
> > and currently it seems to be working fine with dnotify - having a
> > bit of issues getting inotify working with devicemapper ...
> > 
> > I however do not see concern with dnotify also using polling, and
> > would appreciate enlightenment.
> > 
> 
> If possible I would still appreciate points against proposed patch
> (except that it will not apply anymore against cvs :).  If it is the

  Well can you fix it to work against CVS then ? Or at least 0.0.7

> implementation - I am open to suggestions.
> 
> Basically we check each backend in order of preferrence (inotify first,
> as it is a configure option for now), and if it cannot init it, it
> falls through to the next, instead of just checking one.  I am not
> sure how to handle future ones - there might be need for a more
> flexible solution as to what order the backends should be checked,
> or maybe which one first.  Or somehow through gconf?

  In general your patch is about using A vs using B vs. using C, and now
B and C work together and possible A and C too. I just want this to be very 
clear and if you start patching this kind of code you must check you
don't break getting B and C working at the same time.

> Anyhow, if its implementation that bothers, what about attached patch?

  Applied, thanks

gam_server.c: In function `gam_init_subscriptions':
gam_server.c:75: warning: suggest parentheses around assignment used as truth value
gam_server.c:80: warning: suggest parentheses around assignment used as truth value

  please fully parenthesize all boolean expressions in the future please.
I removed ret, it was useless.

> (I am not sure if you would want to use gam_backend_remove_all_for in
> a global instance later on?)

  I'm not sure I fully understand could you be more explicit ?

Daniel

-- 
Daniel Veillard      | Red Hat Desktop team http://redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/